B34A-08 – Applying ILAMB to Data from Several Generations of the Community Land Model to Assess the Relative Contribution of Model Improvements and Forcing Uncertainty to Model–Data Agreement

Authors

David M. Lawrence
National Center for Atmospheric Research
Rosemary Fisher
National Center for Atmospheric Research
Charles Koven
Lawrence Berkeley National Laboratory
Keith W. Oleson
National Center for Atmospheric Research
Sean C. Swenson
National Center for Atmospheric Research
Forrest M. Hoffman (forrest at climatemodeling dot org)
Oak Ridge National Laboratory
James Tremper Randerson
University of California Irvine
Nathan Collier
Oak Ridge National Laboratory
Mingquan Mu
University of California Irvine

Session

Advances in Uncertainty Assessment and Reduction for Terrestrial Carbon Cycle Diagnosis and Prediction II
Wednesday, December 13, 2017 17:45–18:00
New Orleans Ernest N. Morial Convention Center – 383–390

Abstract

The International Land Model Benchmarking (ILAMB) project is a model-data intercomparison and integration project designed to assess and help improve land models. The current package includes assessment of more than 25 land variables across more than 60 global, regional, and site-level (e.g., FLUXNET) datasets. ILAMB employs a broad range of metrics including RMSE, mean error, spatial distributions, interannual variability, and functional relationships. Here, we apply ILAMB for the purpose of assessment of several generations of the Community Land Model (CLM4, CLM4.5, and CLM5). Encouragingly, CLM5, which is the result of model development over the last several years by more than 50 researchers from 15 different institutions, shows broad improvements across many ILAMB metrics including LAI, GPP, vegetation carbon stocks, and the historical net ecosystem carbon balance among others. We will also show that considerable uncertainty arises from the historical climate forcing data used (GSWP3v1 and CRUNCEPv7). ILAMB score variations due to forcing data can be as large for many variables as that due to model structural differences. Strengths and weaknesses and persistent biases across model generations will also be presented.


Forrest M. Hoffman (forrest at climatemodeling dot org)